AgentSkillsCN

making-decisions

当您需要将合成见解转化为具有明确权衡的决策时,可使用此技能。该技能引导您进行互动式决策,并有效识别潜在风险。

SKILL.md
--- frontmatter
name: making-decisions
description: Use when transforming synthesis insights into explicit decisions with documented trade-offs. Guides interactive decision-making and risk identification.
context: fork
agent: general-purpose

Decision Ledger

This skill guides the creation of explicit decisions with full trade-off documentation.

Prerequisites

  • Synthesis complete (/ledger-synthesize)
  • 03-synthesis/CROSS-SYNTHESIS.md exists with decision candidates
  • Per-pillar syntheses in 03-synthesis/SYN-*.md

Workflow

Use TodoWrite to track these mandatory steps:

<required> 1. Load decision candidates from cross-synthesis 2. For each candidate, gather evidence and options 3. Present trade-offs for user decision 4. Create decision entry with semantic ID 5. Identify risks created by each decision 6. Generate DECISIONS.yaml and RISKS.yaml 7. Validate decision quality gates </required>

Step 1: Load Decision Candidates

Read 03-synthesis/CROSS-SYNTHESIS.md to extract:

  • Decision topics needing resolution
  • Supporting evidence from each pillar
  • Identified options

Step 2: Gather Evidence per Decision

For each decision candidate:

  • Collect all relevant EV-* IDs
  • Summarize what evidence supports each option
  • Note confidence levels
yaml
decision_candidate:
  topic: target-segment-priority
  options:
    - name: SMB-first
      evidence:
        - EV-users-smb-pain-points (0.80)
        - EV-economics-smb-unit-economics (0.75)
    - name: Enterprise-first
      evidence:
        - EV-market-enterprise-tam (0.85)
        - EV-competitors-enterprise-gap (0.70)

Step 3: Present Trade-offs

For each decision, present:

  • Options with evidence support
  • What you win with each option
  • What you lose with each option
  • Risks created by each option

Use AskUserQuestion to get user's decision.

Step 4: Create Decision Entry

Write entry to 04-decisions/DECISIONS.yaml using schema from references/decision-ledger-schema.md.

<good-example> ```yaml - id: DEC-scope-smb-first decision: "Target SMB segment before enterprise" status: accepted owner: user created_at: 2026-01-21 alternatives: - Enterprise-first - Multi-segment simultaneous evidence: - EV-users-smb-pain-points - EV-economics-smb-unit-economics - EV-market-enterprise-tam tradeoffs: wins: - "Faster iteration cycles with smaller customers" - "Lower sales friction (self-serve possible)" - "Better unit economics at start" loses: - "Smaller initial contract values" - "May need significant pivot for enterprise later" risks: - RISK-market-smb-churn-rate implications: - "MVP UX must optimize for self-serve onboarding" - "Pricing must fit SMB budget constraints" ``` - Semantic ID (DEC-scope-smb-first) - Explicit alternatives documented - Multiple evidence citations (3+) - Both wins AND loses documented - Linked risks and implications </good-example> <bad-example> ```yaml - id: DEC-001 decision: "Do the thing" status: accepted alternatives: [] evidence: [] tradeoffs: wins: - "It will be good" loses: [] ``` - Non-semantic ID (DEC-001) - Vague decision statement - No alternatives considered - No evidence cited - No loses documented (unrealistic) </bad-example>

Step 5: Identify Created Risks

Each decision may create risks. For each identified risk:

  • Create entry in 05-risks/RISKS.yaml
  • Link back to creating decision
  • Document triggers and mitigations

See references/risk-ledger-schema.md for schema.

Step 6: Generate Ledger Files

Write complete files:

  • 04-decisions/DECISIONS.yaml
  • 05-risks/RISKS.yaml

Step 7: Validate Quality Gates

Decision quality gate:

  • Every decision cites ≥2 evidence IDs
  • Every decision lists ≥1 alternative considered
  • Every decision documents wins AND loses

Risk quality gate:

  • Every risk links to creating decision
  • Every risk has severity and likelihood
  • Every risk has ≥1 mitigation

User Interaction

Use the AskUserQuestion tool for every decision:

Decision prompt

code
Question: "Decision needed: [topic]"
Options:
- "[Option A] - supported by [evidence summary]"
- "[Option B] - supported by [evidence summary]"
- "[Option C] - supported by [evidence summary]"
- "Need more information before deciding"

Trade-off confirmation

code
Question: "You chose [option]. Confirming trade-offs:"
Options:
- "Yes, I accept these trade-offs"
- "Wait, I want to reconsider"
- "Explain the trade-offs more"

Decision status

code
Question: "Should this decision be marked as:"
Options:
- "Accepted (committed)"
- "Provisional (may revisit)"
- "Need more research first"

Risk severity

code
Question: "This decision creates risk: [risk]. How severe?"
Options:
- "High - requires immediate mitigation"
- "Medium - should have mitigation plan"
- "Low - acceptable risk"

Output

After decision-making:

markdown
## Decisions Complete

**Decisions Made:** [N]
**Status:** [X] accepted, [Y] provisional
**Risks Identified:** [Z]

### Decisions Summary
| ID | Decision | Status | Evidence Count |
|----|----------|--------|----------------|
| DEC-scope-smb-first | Target SMB first | accepted | 4 |
| DEC-pricing-freemium | Use freemium model | provisional | 3 |
| ... | ... | ... | ... |

### Risks Created
| ID | Risk | Severity | Linked Decision |
|----|------|----------|-----------------|
| RISK-market-smb-churn | SMB churn rate | medium | DEC-scope-smb-first |
| ... | ... | ... | ... |

### Quality Gate Status
- Decision gate: ✓ All decisions cite ≥2 evidence
- Risk gate: ✓ All risks have mitigations

### Next Step
Run `/ledger-spec` to generate constrained PRD and architecture.

References